Post on 04-Apr-2018
7/29/2019 Hg 3113981406
http://slidepdf.com/reader/full/hg-3113981406 1/9
Md.Saifur Rahman, Nargis Parvin, Md.Tofael Ahmed, Md. Selim Reza, Halida Homyara,
Farhana Enam / International Journal of Engineering Research and Applications (IJERA)
ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.1398-1406
1398 | P a g e
Detection of multiple failures in wavelength division multiplexed
optical network using graph based light path restoration method
Md.Saifur Rahman*, Nargis Parvin**, Md.Tofael Ahmed***, Md. Selim
Reza****, Halida Homyara*****, Farhana Enam*******, *** (Department of Information and Communication Technology, Comilla University, Comilla,Bangladesh)** (Department of Electrical and Electronic Engineering, University of Information Tecnology and Sciences,
Bangladesh)
****, *****, ****** (Department of Information and Communication Engineering, Rajshahi University,Bangladesh)
ABSTRACT
In an Optical WDM network, link/path
failure may occur in a very normal way which
may cause a great amount of data loss if the link is not restored to reroute the affected traffic
accurately and quickly using the redundancy the
network. Wavelength Division Multiplexed(WDM) networks, employing wavelength routing
have emerged as the dominant technology to
satisfy this growing demand for bandwidth. As
the amount of traffic carried has increased, any
single failure can be catastrophic. Survivability
becomes indispensable in such networks.
Therefore, it is imperative to design networks
that can quickly and efficiently recover fromfailures. Most research to date in survivable
optical network design and operation focuses on
single link failures, however, the occurrence of
multiple-link failures are not uncommon in a
network topology. Multi-link failure scenarioscan arise out of two common situations. First, an
arbitrary link may fail in the network, and before
that link can be repaired, another link fails, thus
creating a multi-link failure sequence. Secondly,
it might happen in practice that two distinct
physical links may be routed via the samecommon duct or physical channel. In our
proposed method predefined cycles are used to
survive a failure link. When a link fails at first its
adjacent shorter cycle is used as its backup path.
This can decrease the time complexity. Then
these adjacent cycles are updated by using this
restoration method can survive multiple link
failures. This method theoretically ensures better
performances than existing restoration method.
Keywords – Network survivability, Network
topology, Optical networks, Wavelength Division
Multiplexing (WDM).
1. INTRODUCTIONFailure in a system is a very common
scenario to face. Since networking system is very
much sophisticated, the failure is some what very
natural and high in this case. Failure may occur between two nodes in a transmitted path, again total
path can be failed by any situation. Thus a network
failure in such system may seriously impair servicecontinuity to a countless number of users. So,network survivability is essential. Networks capable
of protecting itself against the failure are calledsurvivable networks [1]. Different restoration
schemes have been proposed to ensure survivability.But an efficient restoration scheme is a demandingrequest in this field.
Our approach is to implement efficient
algorithm for survivability in the case of two or more link failure occurrence. We want to eliminatethe problems in existing restoration scheme and
improve its performance. To do that we propose amethod that can be able- To reduce the timecomplexity in existing restoration scheme when itsearch for a new backup path. To recover a backup path in case of multiple link failure. To reduce theloss of data efficiently when network failure occurs.
The rest of the article is organized as follows.Section 2 provides some description of WDMoptical Network and failures in WDM optical Network have also been discussed. Section 3 provides Existing Restoration Scheme and Their
Drawbacks. Section 4 provides proposedrearrangement based light path restoration schemewith multiple link failure in WDM network. Section
5 provides Description of Experimental Result of Our Proposed Method and Comparison to Other Method and Section 6 concludes the paper.
2. Optical Network, WDM Optical Network
and Failure OccurrenceIn this section, we provide a brief overview
of optical networks, WDM Optical Networks andtheir different generations.
First generation optical networks employfiber only as a transmission medium. Thesenetworks essentially replace copper cables withoptical fibers [2]. The key feature of first generation
optical networks is that all processing is carried outin the electronic domain. The electronics at a nodemust handle all data intended for that node as well
as all data passing through the node and destined toother nodes in the network. Second generationoptical networks use WDM technology to split thehuge bandwidth provided by a fiber into multiple
7/29/2019 Hg 3113981406
http://slidepdf.com/reader/full/hg-3113981406 2/9
Md.Saifur Rahman, Nargis Parvin, Md.Tofael Ahmed, Md. Selim Reza, Halida Homyara,
Farhana Enam / International Journal of Engineering Research and Applications (IJERA)
ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.1398-1406
1399 | P a g e
wavelength channels, that can be used to supportmultiple transmissions simultaneously. Also, someof the switching and routing functions that are
performed by the electronics in first generationoptical network can be carried out in the opticaldomain in second generation optical networks.
Second generation optical networks are over different types of services to the higher network layers. The most commonly used service is thelightpath service. A lightpath [3] is a dedicated
connection on a wavelength between two nodes inthe network, such that no electronic conversiontakes place on the path between these two nodes.
Second generation optical networks employ WDMtechnology, they are also known as WDM networks.A typical WDM link is shown in Fig. 1
Fig. 1 A typical WDM network
It consists of a set of transmitters, opticalamplifiers and receivers. The transmitters are lasers,each supporting one wavelength. The outgoing
signals from different transmitters are multiplexedtogether using a multiplexer. The power amplifier immediately after the multiplexer amplifies the
combined signal. The signal after traveling somedistance on fiber may need amplification again dueto attenuation; this task is carried out by anamplifier. Finally, at the destination, the combinedsignal is amplified again and de-multiplexed. Due todemultiplexing, the signal is split into different
wavelengths which are converted to the electronicdomain using photodetectors, where each photodetector is tuned to a specific wavelength.
In WDM optical networks the basic
following types of failures can be occurred by[4].Link failures: Link failure usually occurs because of cable cuts. When a link fails, all itsconstituents fibers will fail. Node failures: Nodes
failures occur because of equipment failure atnetwork nodes. This type of failure may be the causeof all link failures which are connected to this node.
Fiber failures: A fiber may fail due to the failure of its end components (multiplexers/de-multiplexers)in WXC (Wavelength Cross Connect).Channel
failures: This also possible in WDM networks. Achannel failure is usually caused by the failure of
transmitting and/or receiving equipment operatingon that channel (wavelength).
3. Existing Restoration SchemeIn survivable WDM networks, the lightpath
that carries traffic during normal operations is
known as the primary lightpath. When a primaryLightpath fails, the traffic is rerouted over a newlightpath known as the backup lightpath. In
protection schemes, both primary and backuplightpath are computed before a failure occurs. Inrestoration schemes, a backup lightpath is
discovered dynamically for a primary lightpath after a failure occurs. There similar restoration methodsare may be two categories one is reactive andanother is proactive. The proactive or reactive
schemes can be either link based or path based [5]. The proactive or reactive schemes can be either link based or path based [1]. In link based methodsselect an alternate path between the end nodes of thefailed link. This alternate path along with the intact part of the primary path is used for the recovery. The
method is illustrated in Fig. 2. the figure shows a primary lightpath, A-B-D-G-H, and a backuplightpath, A-B-C-E-D-G-H is used when link B-Dfails. It can be observed that A-B-C-E-D-G-H is
routed around link B-D while retaining the workingsegment of A-B-D-G-H.
Fig. 2 link-based backup path reservationsIn the case of path based methods, a backup path is
computed between the end nodes of the failed primary lightpath. The backup path can use anywavelength independent of the one used by thecorresponding primary lightpath [1]. The path-based
restoration method is illustrated in Figure 3. Figureshows a primary lightpath, A-B-D-G-H, and its
backup lightpath, A-C-E-G-H, on a givenwavelength. Note that A-C-E-G-H is established between the end nodes of A-B-D-G-H, and theworking segment of A-B-D-G-H is not utilized by
A-C-E-G-H.
Fig. 3 path-based backup path reservationsIn the multi-backup method, the FNM is sentdirectly to the source from the first node next to the
failure if it has a link disjoint backup path with thesource and if it has no link disjoint backup path with
the source then it sends this FNM to the node alongthe primary path which has a link disjoint backup
path with the source [6]. In figure 4.8, multi-backup
B
C E G
FD
A H
B
C E G
FD
A H
7/29/2019 Hg 3113981406
http://slidepdf.com/reader/full/hg-3113981406 3/9
7/29/2019 Hg 3113981406
http://slidepdf.com/reader/full/hg-3113981406 4/9
7/29/2019 Hg 3113981406
http://slidepdf.com/reader/full/hg-3113981406 5/9
Md.Saifur Rahman, Nargis Parvin, Md.Tofael Ahmed, Md. Selim Reza, Halida Homyara,
Farhana Enam / International Journal of Engineering Research and Applications (IJERA)
ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.1398-1406
1402 | P a g e
2 u: = t3 while defined u4 insert u to the beginning of S
5 u: = previous[u] Now sequence S is the list of vertices on the shortest path from s to t .
At first we have to determine all theindependent cycle in the network graph G (i, j). Inour approach every node contain all of its adjacent
independent cycle. Only one or two cycle can beadjacent with a link. When a link failure occurs inthe primary path then we determine the independent
cycle (sorter cycle in case of two adjacentindependent cycles) adjacent to that failure link.After a link failure occurs all the nodes of theindependent cycle adjacent to the failure link update
there containing cycle which is adjacent to thatfailure link. (If the failure link has two adjacent
independent cycle then these two cycles create anew cycle otherwise the cycle breaks) Then all other links of that cycle except failure link is used as the backup path of that failure link. The first node next
to the failure link just sends an FNM (Failure Notification Message) to the source then the sourceresends the traffic along with this new path. When
two or more links fail simultaneously (in case of multiple link failure) always active cycle works for surviving iteratively.
All the independent cycles in the network are determined and every node has to store itsadjacent cycles. By using this cycle backup path
nodes adjacent links can be protected.
1. Compute the shortest path, p(l) between node pair (s, d) using Dijkstra’s algorithm for simplicityassume that the primary path consists of a shortest
sequence of nodes.2. If a link failure is detected by p(i) node on the primary path
Then set check =0, Max= number of adjacent cyclesto p(i) node
2.1 For (j=1 to Max)
If p(i)= backuplink (s) and p(i)= backuplink (e)
Then replace the failure link by active path betweenfailure link adjacent nodes in the used cycle.
Set check=1 and break Set length=active backup cycle
length
End loop2.2 if (check=1) Set a new backup path p(m) where
m=l +length
Else No path avail ableEnd If 3. Then the node by which the failure is detected,
send the setup message using the from p(i+m) to
p(1)
4. Then affected traffics are retransmitted along theselected new backup path.5. Then all the nodes on adjacent cycle to the failure
link just update their affected cycle.6. If another failure is detected again start from step2 to establish a new backup path.
We have implemented our proposedalgorithm by C language. We have tested our implemented algorithm or program for the various
graphs. In our implementation at first we determineall the independent cycles of the network using thelink adjacent matrix and the cost matrix. In adjacent
matrix ije =1 for i ≠ j and ij
e belongs to E . Where E
is a set of edges otherwise ije =0. In cost matrix
ijc =weight for i ≠ j and ij
c belongs to E . Where E
is a set of edges otherwise ijc =0. Then all the cycles
are determined from the matrix are as above figure.
Fig.9 adjacent or path matrix and cost matrix
Where first chart represents the cycles before failure and its first column represents the
number of cycle’s node. Second chart represents thecycles after two or more failures and its first columnrepresents the number of cycle’s node. Then byusing first chart’s cycles we can establish the backup
path for the single failure link and rearrange them assecond chart, we can establish the backup path for the multiple failure links. In the following figure
independent cycles before failure and initial primary path, Independent cycles after one link failure and backup path, and Independent cycles after two links
failure and corresponding backup path are shownfrom implementing output.
11e
12e
13e
ne1
21e
22e 23
en
e2
1ne
nne
……
……
7/29/2019 Hg 3113981406
http://slidepdf.com/reader/full/hg-3113981406 6/9
7/29/2019 Hg 3113981406
http://slidepdf.com/reader/full/hg-3113981406 7/9
Md.Saifur Rahman, Nargis Parvin, Md.Tofael Ahmed, Md. Selim Reza, Halida Homyara,
Farhana Enam / International Journal of Engineering Research and Applications (IJERA)
ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.1398-1406
1404 | P a g e
second so for 11-16 link failure 11-16-17-11 is usedas protection cycle of 11-16 link. Then 16 send anFNM to A using 16-17-11-6-1 and 1 resends the
traffic to 21 using 1-6-11-17-16-21. Then 11, 10, 15,16 and 17 nodes update their cycle which contains11-16 links. As 11 update its 11-16-17-11 or 11-16-
15-10-10 by 11-10-15-16-17 Cycle. . This methodneeds 4 nodes to recover 11-16 links after fails.
3
2 19
1
20
21
10
11 16
12
6
5
7 17
15
Retransmit
traffic
FNM
Primary path
fail
3
2 19
1
20
21
10
11 16
12
6
5
7 17
15
Retransmit
traffic
FNM
Primary path
fail
17
1611
Cycle 16-17-11-16
17
1611
17
1611
Cycle 16-17-11-16
Fig. 14 restoration of backup path for a Singlefailure link by independent cycle in our proposedmethod.
5.2 Description of Experimental Result with
Multiple Links FailureTo recover multiple failures in the
following network, we have considered source 1 anddestination 12 then the primary path is 1-2-5-6-9-12is selected. When first failure occurs at link 2-5 then
according to “ Dual-Link Failure Resiliency through Backup Link Mutual Exclusion” method backup pathwill be 1-4-7-10-13-12 (in case of “ Link Protection
Failure Independent Protection (LP-FIP)”). In this
case backup path is mutually exclusive to the primary path by Chandak et al [2]. Now if 4-7 link fails there is no any mutually exclusive path then the
network is block for the affected traffics. Becausethere is no any mutual exclusive backup path for 4-7link
(a)
(b)
Fig. 15 multiple failure restoration using (a) existingdual-link failure resiliency through backup link mutual exclusion method (b) proposed method
In our proposed method, the primary path is 1-2-5-6-9-12.When the link 2-5 fails, then the back up path1-4-5-6-9-12 is established by using the cycle 5-4-1-
2-5. When second failure occurs at the link 5-6, thenthe back up path 1-4-7-6-9-12 is established byusing the cycle 6-7-4-5-6.
5.2.1 Comparison Result with Other Existing
MethodTo compare the performance of our
proposed, we have compared our proposed methodwith existing path based method, multi backupmethod and rearrangement based method in case of single link failure and with LP-FIP method in case
of multiple links failure.
5.2.2 Comparison Results with existing methodsin case of single link failure
In figure: 11, 12, 13 and 14 when link 11-16 fails then the number of nodes required to restore
the backup path, selecting path to notify the failureand the new restored backup path are as followingTable I. from this table we can show that to restore a
backup path to notify the failure to source, path based method needs 8 nodes, Multi Backup Methodneeds 6 nodes, existing rearrangement based methodneeds 5 nodes and Proposed Method needs only 4nodes. So our proposed method reduced the number of nodes to transmit failure notification to the source
node from first node next to the failure link.Table I. Comparison result for failure link recoveryfrom above four figures.
RestorationScheme
No of nodesto
Recover path
Nodes UsedtoAcknowledge
Nodes UsedtoResend Data
Path Based
Method
8 16-21-19-15-
10-5-2-1
1-2-5-10-15-
19-21
MultiBackup
Method
6 16-15-10-5-2-1
1-2-5-10-15-16-21
Rearrangement Based
5 16-17-11-6-1 1-6-11-17-16-21
ProposedMethod
4 16-17-11 1-6-11-17-16-21
The experimental results in the following Table IIshow the required average numbers of nodes torecover a single link failure over the total number of
network nodes and links. In fig. 13 we have
7/29/2019 Hg 3113981406
http://slidepdf.com/reader/full/hg-3113981406 8/9
Md.Saifur Rahman, Nargis Parvin, Md.Tofael Ahmed, Md. Selim Reza, Halida Homyara,
Farhana Enam / International Journal of Engineering Research and Applications (IJERA)
ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.1398-1406
1405 | P a g e
compared our proposed method with path basedmethod, multibackup method and existingrearrangement based method in case of single link
failure recovery.Table II Average length of backup paths in variousnetworks (In case single link failure)
Network No
1 2 3 4 5
No of hopes
12 21 27
45 65
No of
Links
16 25 5
0
64 83
Required No of nodes/hopes for restoring backup path
By path basedmethod
6 7 9 13 17
By multi
backupmethod
5 6 7 11 13
Byrearrangement based
3 4 6 8 11
By proposedmethod
3 4 6 8 11
Fig. 16 Average numbers of nodes required torestore backup paths by existing method and
proposed method
Table I. shows that the multi backup method canreduce the number of hopes to recover a failure link
than path based method. On the other hand therearrangement based method can show better performance than multi backup method.
5.2.3 Comparison Results with Existing
Restoration Scheme in Case of Multiple FailuresThe experimental results in the following
table show the required number of nodes to recover multiple links failure over the total number of network nodes and network links.
In fig. 17 we have compared our proposed method
with LP-FIP in Dual-link failure resiliency along backup link mutual exclusion in case of single link failure and in case of multi links failure.
Table III Average length of backup paths in variousnetworks (With multiple links failure)
Network No 1 2 3 4 5
No of hopes 12 21 27 45 65
No of Links 16 25 50 64 83
Required No of nodes/hopes for restoring backup
pathLP-FIPmethod
Singlefail
3 5 6 9 13
Multifails
8 11 15 23 31
Proposed
method
Single
fail
3 5 6 9 13
Multifails
6 9 13 17 23
Fig. 17 average numbers of nodes required to restore backup paths by LP-FIP and proposed method
5.3 Complexity AnalysisWe can express the running time of
Dijkstra's algorithm on a graph with m edges and nvertices as a function of m and n using the Big Onotation. Each iteration, need n*(n+1)/2comparisons so complexity in this algorithm to find
shortest path is O(2
n ). In our proposed method to
select primary path Dijkstra's algorithm is used. So
complexity to select primary path is O( 2
n ). Now if
n is the total the number nodes or hops and m is totalthe number of links or edges aaccording to our proposed method the network has m-(n-1) number
of independent cycles in the network where everyindependent cycle has maximum n nodes. By usingthese cycles we can determine m (n<m) backup pathagainst every link. So we total comparison for finding backup path m where n=m or n<m. So thetime complexity to search a backup cycle is O (n).
0
20
40
1 2 3 4 5 R e q u i r e d h o p s
Networks
Comparison Chart between existing and
proposed method Existingmethod(sin
gle failure)
Existingmethod(mu
ltiplefailure)
0
5
10
15
20
1 2 3 4 5
R e q u i r e d h o p s
Networks
Compare performance between existing
and proposed method to resore pathPathbased
method
Multibackup
method
7/29/2019 Hg 3113981406
http://slidepdf.com/reader/full/hg-3113981406 9/9
Md.Saifur Rahman, Nargis Parvin, Md.Tofael Ahmed, Md. Selim Reza, Halida Homyara,
Farhana Enam / International Journal of Engineering Research and Applications (IJERA)
ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.1398-1406
1406 | P a g e
Table IV. Comparison of Complexity with variousexisting restoration schemes
No
Restorationmethod to
survive
Searching
algorithm to
search backup path
Complexity for
primary path
Complexity to
restore path
1 Path based Used tosearchfull path
O(2
n ) O(2
n )
2 Multi backup Used to
searchfull path
O(2
n ) O(2
n )
3 Rearrangement
Used tosearchfull path
O(2
n ) O(2
n )
4 Proposed
method
Search
only backupcycle
O(
2
n )
O(n)
From complexity analysis we show that to select primary path our proposed method uses Dijkstra’salgorithm as other method has the complexity O(
2n ). But to restore the backup path existing
methods use Dijkstra’s algorithm so the complexity
to restore backup path is O( 2
n ). But to restore the
backup path for the failure link proposed method useonly predefined independent cycle adjacent to thefailure link which reduces the time complexity. In
our proposed method the complexity is O( n ) torestore a backup path. So our proposed method
significantly works fast to restore backup path.
6. Concluding RemarksThe current multi-backup method performs
better than the path based method but rearrangementmethod shows a better performance against themulti-backup method. Which reduces the backup path length, and hence the restoration time is alsoreduced. But to restore the backup path for thefailure link this method uses Dijkstra’s algorithm
that increases the time complexity of this process. Ina large network multiple failures are not uncommon, but this method can not handle the multiple failures.In our proposed method, the required nodes to
restore a failure link are same. But to restore the backup path for the failure link proposed method useonly shortest predefined independent cycles,
adjacent to the failure link which reduces the timecomplexity of this method. After restore a backup path independent cycles adjacent to the failure link are updated that make this method able to handle
multiple failures. So this method shows better performance than existing rearrangement basedmethod. In case of multiple failures our proposed
method can solve the blocking problem in LP-FIP.Our proposed method has some limitations, such as:
When the failures of site equipments or nodefailures create multiple links failure in a network then this method fail to survive.
This method is faster than existingrearrangement based method to survive single link failure. But after survive the first failure, to update
the adjacent cycles to the failure link takes sometime that may slow to restore backup path of second,third or more other failure links.
To show better performance we have the
following recommendation for our proposedmethod, to reduce the limitations of our proposedmethod the network components must be reliable,
and have to ability to provide backup.
R EFERENCES [1] Sateesh Chandra Shekhar, “Survivable
Multicasting in WDM Optical Networks ,”By Bachelor of Engineering (Computer
Engineering) University of Pune, August2004.
[2] O. Crochat, J. Y. Boudec, and O.Gerstel,“ Protection interoperability for
WDM optical networks," IEEE/ACMTransactions on Networking, June 2000.
[3] Pallab Datta, “Survivability approaches for
multiple failures in WDM optical networks,” Iowa State University Ames,Iowa 2005.
[3] C.Silva Ram Murthy and MohanGurusami,”WDM Optical NetworksConcepts, Design, and Algorithm,”
Pearsonal Education.[4] Andrew Tanenbaum “Computer
Networks”-fourth edition Vrije Universiteit,Amsterdam, The Netherlands 2007-2008.
[5] Son Hong Ngo,” Dynamic Lightpath Protection in WDM optical Networks Using Ant-based Mobile Agents,” GraduateSchool of Information Science JAIST,
Japan Email: sonhong@jaist.ac.jp November 2005.
[6] M. M. A. Azim and Md. Mamun-ur-Rashid
Khandker, “ An Active Lightpath Restoration Method Using Multi-Backup
Paths in WDM Networks”, Proceedings of the International Symposium on Towards
Peta-Bit Ultra-Networks, Ishikawa, Japan,September 8-9, 2003.
[7] Md.Mahabubur Rahman, Syed Md.Galib
and Md.Hasem Uddin,”a new Rearrangement based Lightpath Restoration Scheme in Optical Networks”,
International Conference on Computer andInformation Technology (ICCIT)’2005.
[8] Retrieved from
http://en.giswiki.net/wiki/Dijkstra%27s_algorithm